Global energy transition under dynamic trade networks: Integrating complex-network and deep learning methods

The rapid advancement of energy technologies and profound geopolitical shifts have intensified uncertainties in global energy supply chains and added complexity to energy transitions. Existing transitions indicators, however, often fail to capture the evolution of international energy trade networks and the multidimensional interactions among countries embedded within them. To address this gap, this study integrates conventional transition indicators, weighted complex network analysis, and deep learning to jointly account for domestic energy system changes and external trade dynamics, thereby enabling a more comprehensive assessment of national energy transition performance. Specifically, a five-dimensional indicator framework is developed, encompassing energy security, energy equity, energy sustainability, transition support, and transition cost. The bilateral energy trade network of 115 countries are simulated for 2015-2023 in terms of scale, efficiency, and polarization. A Relational Graph Convolutional Network (R-GCN) is employed to integrate node-level attributes and relational features within the global energy trade network, producing a composite energy transition index and revealing cross-country trends. Results indicate that the global energy transition index exhibits a generally upward yet fluctuating trajectory, while transition support declines. Network centrality in global energy trade has risen by approximately 70%, with hub countries achieving sustained improvements, though the transition gap with lagging regions has widened by nearly 27%. Sensitivity to trade shocks varies substantially across dimensions, with energy sustainability being least affected. These findings underscore the need for countries to align transition strategies with regional trade network structures to improve both the efficiency and resilience of the global energy transition.